What is Spikefrost
Spikefrost is a platform where you build real software by talking to an AI agent — and where that software runs. What it gives you, who it's for, and how the pieces fit together.
Spikefrost is a platform for building and running software by describing what you want. You talk to an AI agent; it designs, writes, and deploys a working application. The application is not a mockup or a prototype — it is a real deployed app with a URL, a database, and file storage, running on a global edge network from the first minute.
There are two ways in, and they operate on exactly the same app:
- The desktop app — a workspace where you chat with an agent that builds for you. No terminal, no code required. See The desktop app.
- The
sfCLI — the developer interface. Open the project in your own editor, run your own coding agent (Claude Code or Codex), deploy from a terminal. See The sf CLI.
Neither is a lesser version of the other. The same files, the same history, the same deploys.
What you get with every app
Creating an app provisions a full stack. Nothing here is an add-on you wire up later:
| Capability | What it means |
|---|---|
| Edge web app | A Cloudflare Worker serving your app worldwide, with its own URL immediately |
| SQL database | A D1 (SQLite) database with versioned migrations and point-in-time recovery |
| Key-value store | A KV namespace for caching, sessions, and flags |
| File storage + CDN | Public assets served from /assets/..., plus private artifact storage |
| Version control | Every AI turn and every edit lands as a commit, with branches and rollback |
| Environments | Production plus any number of named environments (staging, dev) inside the same app |
| Custom domains | Attach your own domain with automatic TLS |
| Observability | Request logs, per-app metrics, and a per-deploy event stream |
| AI agents | Optional agents that live in the app's code and act through its channels |
The idea behind it
Most tools make you choose between fast and real. Prototype builders are fast and produce something you eventually throw away. Traditional development is real and takes a quarter. Spikefrost's bet is that an AI agent working inside a platform that already owns the boring parts — hosting, database, deploys, secrets, domains, versioning — collapses that trade-off. The agent does not have to invent infrastructure. It writes application code against bindings that already exist, and ships.
That has a consequence worth stating plainly: your app is a codebase, not a saved configuration. You can read it, a developer can review it, and it behaves like software rather than like a settings screen.
Apps versus agents
Two words that get conflated everywhere. On Spikefrost they are distinct:
- An app is the deployed thing: routes, pages, data, storage. A landing page is an app. A dashboard is an app. A store is an app.
- An agent is an actor that lives inside an app. It has instructions, a model, tools it is allowed to use, and channels it can be reached on. It can answer an email, watch a queue, run on a schedule, or hold a conversation in a chat widget.
An app with no agents is a normal web app that happens to have been written by one. An app with agents is what we call an agentic app: software that both serves requests and takes action. Most real apps end up as a bit of both — a website with a support agent behind the chat widget, or an internal tool with an agent that triages the overnight queue.
Who uses it, and for what
- Founders and operators building the thing they need this week without hiring for it — booking pages, internal tools, customer portals, landing pages with a real chat.
- Developers who want to skip the setup tax. The stack is opinionated on purpose: Hono, Workers, D1. You get a deployed app on day one instead of a
terraform plan. - Teams replacing manual back-office work with agents that have real credentials, real audit trails, and real limits — with permissions and spend visible to whoever owns the team.
What Spikefrost is not
Being clear about the edges saves you time:
- Not a general VPS or container host. Apps run on Workers at the edge. That is what makes them fast and global; it also means long-running processes, arbitrary binaries, and native daemons are the wrong shape. Long or heavy work belongs in a scheduled routine, a queue consumer, or an agent turn.
- Not a design tool. Agents produce a real visual identity, but this is not Figma.
- Not a place to store bulk files. Your project tree is for source. Media goes to assets, data goes in the database — see Files and assets.
Where to go next
- Getting started — install, sign in, build and deploy your first app.
- Core concepts — the vocabulary: apps, environments, agents, connectors, jobs, commits.
- App anatomy — for developers: what the project tree looks like and what's bound where.
- Guides — longer-form material on designing and running agents.
Frequently asked questions
Is Spikefrost a chatbot builder?
No. A chatbot answers questions. A Spikefrost app is real deployed software — a web app with its own database, file storage, scheduled work, and optional AI agents that act on your behalf. A chat widget is one thing you can add to it, not the product.
Do I need to be a developer?
No, and being one still helps. Non-technical users build entirely through conversation in the desktop app. Developers get a normal codebase, a CLI, version control, and multiple environments — the same app, opened from the other end.
Where does my app actually run?
On a global edge network. Each app is a Cloudflare Worker with its own SQL database (D1), key-value store, file storage, and CDN-backed assets, deployed to every edge location at once.
What does an app cost to run?
Hosting an app on Spikefrost is part of your plan; there is no per-app server to size or pay for. AI model usage is metered against your team balance, so the cost that varies is how much you use agents — not how much traffic your site gets.
Can I take my code with me?
Yes. An app is a normal TypeScript project — Hono on Cloudflare Workers, SQL migrations, standard tooling. You can read every file, edit it in your own editor, and copy it out.